arXiv subjects

Abhik Ghosh

Publications and source records attributed to Abhik Ghosh.

At least 55 records · Page 3Linked to original sources

arXiv2023

Robust adaptive Lasso in high-dimensional logistic regression

Penalized logistic regression is extremely useful for binary classification with large number of covariates (higher than the sample size), having several real life applications, including genomic disease classification. However, the existing methods based on the likelihood loss function are sensitive to data contamination and other noise and, hence, robust methods are needed for stable and more accurate inference. In this paper, we propose a family of robust estimators for sparse logistic models utilizing the popular density power divergence based loss function and the general adaptively weighted LASSO penalties. We study the local robustness of the proposed estimators through its influence function and also derive its oracle properties and asymptotic distribution. With extensive empirical illustrations, we demonstrate the significantly improved performance of our proposed estimators over the existing ones with particular gain in robustness. Our proposal is finally applied to analyse four different real datasets for cancer classification, obtaining robust and accurate models, that simultaneously performs gene selection and patient classification.

stat.ME↗

arXiv2023

Exponential Consistency of M-estimators in Generalized Linear Mixed Models

Generalized linear mixed models are powerful tools for analyzing clustered data, where the unknown parameters are classically (and most commonly) estimated by the maximum likelihood and restricted maximum likelihood procedures. However, since the likelihood based procedures are known to be highly sensitive to outliers, M-estimators have become popular as a means to obtain robust estimates under possible data contamination. In this paper, we prove that, for sufficiently smooth general loss functions defining the M-estimators in generalized linear mixed models, the tail probability of the deviation between the estimated and the true regression coefficients have an exponential bound. This implies an exponential rate of consistency of these M-estimators under appropriate assumptions, generalizing the existing exponential consistency results from univariate to multivariate responses. We have illustrated this theoretical result further for the special examples of the maximum likelihood estimator and the robust minimum density power divergence estimator, a popular example of model-based M-estimators, in the settings of linear and logistic mixed models, comparing it with the empirical rate of convergence through simulation studies.

math.ST↗

arXiv2023

Existence and Consistency of the Maximum Pseudo \b{eta}-Likelihood Estimators for Multivariate Normal Mixture Models

Robust estimation under multivariate normal (MVN) mixture model is always a computational challenge. A recently proposed maximum pseudo \b{eta}-likelihood estimator aims to estimate the unknown parameters of a MVN mixture model in the spirit of minimum density power divergence (DPD) methodology but with a relatively simpler and tractable computational algorithm even for larger dimensions. In this letter, we will rigorously derive the existence and weak consistency of the maximum pseudo \b{eta}-likelihood estimator in case of MVN mixture models under a reasonable set of assumptions.

math.ST↗

arXiv2023

Towards $21$-cm intensity mapping at $z=2.28$ with uGMRT using the tapered gridded estimator II: Cross-polarization power spectrum

Neutral hydrogen ($\rm{HI}$) $21$-cm intensity mapping (IM) offers an efficient technique for mapping the large-scale structures in the universe. We introduce the 'Cross' Tapered Gridded Estimator (Cross TGE), which cross-correlates two cross-polarizations (RR and LL) to estimate the multi-frequency angular power spectrum (MAPS) $C_{\ell}(Δν)$. We expect this to mitigate several effects like noise bias, calibration errors etc., which affect the 'Total' TGE which combines the two polarizations. Here we apply the Cross TGE on a $24.4 \,\rm{MHz}$ bandwidth uGMRT Band $3$ data centred at $432.8 \,\rm{MHz}$ aiming $\rm{HI}$ IM at $z=2.28$. The measured $C_{\ell}(Δν)$ is modelled to yield maximum likelihood estimates of the foregrounds and the spherical power spectrum $P(k)$ in several $k$ bins. Considering the mean squared brightness temperature fluctuations, we report a $2σ$ upper limit $Δ_{UL}^{2}(k) \le (58.67)^{2} \, {\rm mK}^{2}$ at $k=0.804 \, {\rm Mpc}^{-1}$ which is a factor of $5.2$ improvement on our previous estimate based on the Total TGE. Assuming that the $\rm{HI}$ traces the underlying matter distribution, we have modelled $C_{\ell}(Δν)$ to simultaneously estimate the foregrounds and $[Ω_{\rm{HI}} b_{\rm{HI}}] $ where $Ω_{\rm{HI}}$ and $b_{\rm{HI}}$ are the $\rm{HI}$ density and linear bias parameters respectively. We obtain a best fit value of $[Ω_{\rm{HI}}b_{\rm{HI}}]^2 = 7.51\times 10^{-4} \pm 1.47\times 10^{-3}$ which is consistent with noise. Although the $2σ$ upper limit $[Ω_{\rm{HI}}b_{\rm{HI}}]_{UL} \leq 0.061$ is $\sim 50$ times larger than the expected value, this is a considerable improvement over earlier works at this redshift.

astro-ph.CO↗

arXiv2022

Robust Estimation of Average Treatment Effects from Panel Data

In order to evaluate the impact of a policy intervention on a group of units over time, it is important to correctly estimate the average treatment effect (ATE) measure. Due to lack of robustness of the existing procedures of estimating ATE from panel data, in this paper, we introduce a robust estimator of the ATE and the subsequent inference procedures using the popular approach of minimum density power divergence inference. Asymptotic properties of the proposed ATE estimator are derived and used to construct robust test statistics for testing parametric hypotheses related to the ATE. Besides asymptotic analyses of efficiency and powers, extensive simulation studies are conducted to study the finite-sample performances of our proposed estimation and testing procedures under both pure and contaminated data. The robustness of the ATE estimator is further investigated theoretically through the influence functions analyses. Finally our proposal is applied to study the long-term economic effects of the 2004 Indian Ocean earthquake and tsunami on the (per-capita) gross domestic products (GDP) of five mostly affected countries, namely Indonesia, Sri Lanka, Thailand, India and Maldives.

stat.ME↗

arXiv2022

Robust Sure Independence Screening for Non-polynomial dimensional Generalized Linear Models

We consider the problem of variable screening in ultra-high dimensional generalized linear models (GLMs) of non-polynomial orders. Since the popular SIS approach is extremely unstable in the presence of contamination and noise, we discuss a new robust screening procedure based on the minimum density power divergence estimator (MDPDE) of the marginal regression coefficients. Our proposed screening procedure performs well under pure and contaminated data scenarios. We provide a theoretical motivation for the use of marginal MDPDEs for variable screening from both population as well as sample aspects; in particular, we prove that the marginal MDPDEs are uniformly consistent leading to the sure screening property of our proposed algorithm. Finally, we propose an appropriate MDPDE based extension for robust conditional screening in GLMs along with the derivation of its sure screening property. Our proposed methods are illustrated through extensive numerical studies along with an interesting real data application.

math.ST↗

arXiv2022

Probing early universe through redshifted 21-cm signal: Modelling and observational challenges

The redshifted 21-cm radiation from the atomic hydrogen (HI) provides an excellent direct probe to study the evolution of HI in IGM and thus reveal the nature of the first luminous objects, their evolution and role during Cosmic Dawn (CD) and Epoch of Reionization (EoR), and formation and evolution of the structures thereafter. Direct mapping of the HI density during the CD-EoR is rather difficult with the current and future instruments due to large foreground and other observational contamination. The first detection of this redshifted HI signal is planned through statistical estimators. Given the utmost importance of the detection and analysis of the redshifted 21-cm signal, physics of CD-EoR is one of the objectives of the upcoming SKA-Low telescope. This paper summarizes the collective effort of Indian astronomers to understand the origin of the redshifted 21-cm signal, sources of first ionizing photons, their propagation through the IGM, various cosmological effects on the expected 21-cm signal, various statistical measures of the signal like power spectrum, bispectrum, etc. A collective effort on detection of the signal by developing estimators of the statistical measures with rigorous assessment of their expected uncertainties, various challenges like that of the large foreground emission and calibration issues are also discussed. Various versions of the detection methods discussed here have also been used in practice with the GMRT with successful assessment of the foreground contamination and upper limits on the matter density in EoR and post-EoR. The collective efforts compiled here has been a large part of the global effort to prepare proper observational technique and analysis procedure for the first light of the CD-EoR through the SKA-Low.

astro-ph.CO↗

arXiv2022

Towards 21-cm Intensity Mapping at $z=2.28$ with uGMRT using the Tapered Gridded Estimator I: Foreground Avoidance

The post-reionization $(z \le 6)$ neutral hydrogen (HI) 21-cm intensity mapping signal holds the potential to probe the large scale structures, study the expansion history and constrain various cosmological parameters. Here we apply the Tapered Gridded Estimator (TGE) to estimate $P(k_{\perp},k_{\parallel})$ the power spectrum of the $z = 2.28$ $(432.8\, {\rm MHz})$ redshifted 21-cm signal using a $24.4\,{\rm MHz}$ sub-band drawn from uGMRT Band 3 observations of European Large-Area ISO Survey-North 1 (ELAIS-N1). The TGE allows us to taper the sky response which suppresses the foreground contribution from sources in the periphery of the telescope's field of view. We apply the TGE on the measured visibility data to estimate the multi-frequency angular power spectrum (MAPS) $C_{\ell}(Δν)$ from which we determine $P(k_{\perp},k_{\parallel})$ using maximum-likelihood which naturally overcomes the issue of missing frequency channels (55 \% here). The entire methodology is validated using simulations. For the data, using the foreground avoidance technique, we obtain a $2\,σ$ upper limit of $Δ^2(k) \le (133.97)^2 \, {\rm mK}^{2}$ for the 21-cm brightness temperature fluctuation at $k = 0.347 \, \textrm{Mpc}^{-1}$. This corresponds to $[Ω_{\rm HI}b_{\rm HI}] \le 0.23$, where $Ω_{\rm HI}$ and $b_{\rm HI}$ respectively denote the cosmic \HI mass density and the \HI bias parameter. A previous work has analyzed $8 \, {\rm MHz}$ of the same data at $z=2.19$, and reported $Δ^{2}(k) \le (61.49)^{2} \, {\rm mK}^{2}$ and $[Ω_{\rm HI} b_{\rm HI}] \le 0.11$ at $k=1 \, {\rm Mpc}^{-1}$. The upper limits presented here are still orders of magnitude larger than the expected signal corresponding to $Ω_{\rm HI} \sim 10^{-3}$ and $b_{\rm HI} \sim 2 $.

astro-ph.CO↗

arXiv2022

Robust and Efficient Parameter Estimation for Discretely Observed Stochastic Processes

In various practical situations, we encounter data from stochastic processes which can be efficiently modelled by an appropriate parametric model for subsequent statistical analyses. Unfortunately, the most common estimation and inference methods based on the maximum likelihood (ML) principle are susceptible to minor deviations from assumed model or data contamination due to their well known lack of robustness. Since the alternative non-parametric procedures often lose significant efficiency, in this paper, we develop a robust parameter estimation procedure for discretely observed data from a parametric stochastic process model which exploits the nice properties of the popular density power divergence measure in the framework of minimum distance inference. In particular, here we define the minimum density power divergence estimators (MDPDE) for the independent increment and the Markov processes. We establish the asymptotic consistency and distributional results for the proposed MDPDEs in these dependent stochastic process set-ups and illustrate their benefits over the usual ML estimator for common examples like Poisson process, drifted Brownian motion and auto-regressive models.

stat.ME↗

arXiv2021

A Scale-invariant Generalization of the Rényi Entropy, Associated Divergences and their Optimizations under Tsallis' Nonextensive Framework

Entropy and relative or cross entropy measures are two very fundamental concepts in information theory and are also widely used for statistical inference across disciplines. The related optimization problems, in particular the maximization of the entropy and the minimization of the cross entropy or relative entropy (divergence), are essential for general logical inference in our physical world. In this paper, we discuss a two parameter generalization of the popular Rényi entropy and associated optimization problems. We derive the desired entropic characteristics of the new generalized entropy measure including its positivity, expandability, extensivity and generalized (sub-)additivity. More importantly, when considered over the class of sub-probabilities, our new family turns out to be scale-invariant. We also propose the corresponding cross entropy and relative entropy measures and discuss their geometric properties including generalized Pythagorean results over $β$-convex sets. The maximization of the new entropy and the minimization of the corresponding cross or relative entropy measures are carried out explicitly under the non-extensive (`third-choice') constraints given by the Tsallis' normalized $q$ expectations which also correspond to the $β$-linear family of probability distributions. Important properties of the associated forward and reverse projection rules are discussed along with their existence and uniqueness. In this context, we have come up with, for the first time, a class of entropy measures -- a subfamily of our two-parameter generalization -- that leads to the classical (extensive) exponential family of MaxEnt distributions under the non-extensive constraints. Other members of the new entropy family, however, lead to the power-law type generalized $q$-exponential MaxEnt distributions which is in conformity with Tsallis' nonextensive theory.

math.ST↗

arXiv2021

Robust Parametric Inference for Finite Markov Chains

We consider the problem of statistical inference in a parametric finite Markov chain model and develop a robust estimator of the parameters defining the transition probabilities via minimization of a suitable (empirical) version of the popular density power divergence. Based on a long sequence of observations from a first-order stationary Markov chain, we have defined the minimum density power divergence estimator (MDPDE) of the underlying parameter and rigorously derived its asymptotic and robustness properties under appropriate conditions. Performance of the MDPDEs is illustrated theoretically as well as empirically for some common examples of finite Markov chain models. Its applications in robust testing of statistical hypotheses are also discussed along with (parametric) comparison of two Markov chain sequences. Several directions for extending the MDPDE and related inference are also briefly discussed for multiple sequences of Markov chains, higher order Markov chains and non-stationary Markov chains with time-dependent transition probabilities. Finally, our proposal is applied to analyze corporate credit rating migration data of three international markets.

stat.ME↗

arXiv2021

Strata-based Quantification of Distributional Uncertainty in Socio-Economic Indicators: A Comparative Study of Indian States

This paper reports a comprehensive study of distributional uncertainty in a few socio-economic indicators across the various states of India over the years 2001-2011. We show that the DGB distribution, a typical rank order distribution, provide excellent fits to the district-wise empirical data for the population size, literacy rate (LR) and work participation rate (WPR) within every states in India, through its two distributional parameters. Moreover, taking resort to the entropy formulation of the DGB distribution, a proposed uncertainty percentage (UP) unveils the dynamics of the uncertainty of LR and WPR in all states of India. We have also commented on the changes in the estimated parameters and the UP values from the years 2001 to 2011. Additionally, a gender based analysis of the distribution of these important socio-economic variables within different states of India has also been discussed. Interestingly, it has been observed that, although the distributions of the numbers of literate and working people has a direct (linear) correspondence with that of the population size, the literacy and work-participation rates are distributed independently of the population distributions.

stat.AP↗

arXiv2021

RaJIVE: Robust Angle Based JIVE for Integrating Noisy Multi-Source Data

With increasing availability of high dimensional, multi-source data, the identification of joint and data specific patterns of variability has become a subject of interest in many research areas. Several matrix decomposition methods have been formulated for this purpose, for example JIVE (Joint and Individual Variation Explained), and its angle based variation, aJIVE. Although the effect of data contamination on the estimated joint and individual components has not been considered in the literature, gross errors and outliers in the data can cause instability in such methods, and lead to incorrect estimation of joint and individual variance components. We focus on the aJIVE factorization method and provide a thorough analysis of the effect outliers on the resulting variation decomposition. After showing that such effect is not negligible when all data-sources are contaminated, we propose a robust extension of aJIVE (RaJIVE) that integrates a robust formulation of the singular value decomposition into the aJIVE approach. The proposed RaJIVE is shown to provide correct decompositions even in the presence of outliers and improves the performance of aJIVE. We use extensive simulation studies with different levels of data contamination to compare the two methods. Finally, we describe an application of RaJIVE to a multi-omics breast cancer dataset from The Cancer Genome Atlas. We provide the R package RaJIVE with a ready-to-use implementation of the methods and documentation of code and examples.

stat.ME↗

arXiv2020

Demonstrating the Tapered Gridded Estimator (TGE) for the Cosmological HI 21-cm Power Spectrum using $150 \, {\rm MHz}$ GMRT observations

We apply the Tapered Gridded Estimator (TGE) for estimating the cosmological 21-cm power spectrum from $150 \, {\rm MHz}$ GMRT observations which corresponds to the neutral hydrogen (HI) at redshift $z = 8.28$. Here TGE is used to measure the Multi-frequency Angular Power Spectrum (MAPS) $C_{\ell}(Δν)$ first, from which we estimate the 21-cm power spectrum $P(k_{\perp},k_{\parallel})$. The data here are much too small for a detection, and the aim is to demonstrate the capabilities of the estimator. We find that the estimated power spectrum is consistent with the expected foreground and noise behaviour. This demonstrates that this estimator correctly estimates the noise bias and subtracts this out to yield an unbiased estimate of the power spectrum. More than $47\%$ of the frequency channels had to be discarded from the data owing to radio-frequency interference, however the estimated power spectrum does not show any artifacts due to missing channels. Finally, we show that it is possible to suppress the foreground contribution by tapering the sky response at large angular separations from the phase center. We combine the k modes within a rectangular region in the `EoR window' to obtain the spherically binned averaged dimensionless power spectra $Δ^{2}(k)$ along with the statistical error $σ$ associated with the measured $Δ^{2}(k)$. The lowest $k$-bin yields $Δ^{2}(k)=(61.47)^{2}\,{\rm K}^{2}$ at $k=1.59\,\textrm{Mpc}^{-1}$, with $σ=(27.40)^{2} \, {\rm K}^{2}$. We obtain a $2 \, σ$ upper limit of $(72.66)^{2}\,\textrm{K}^{2}$ on the mean squared HI 21-cm brightness temperature fluctuations at $k=1.59\,\textrm{Mpc}^{-1}$.

astro-ph.CO↗

arXiv2020

Robust Estimation under Linear Mixed Models: The Minimum Density Power Divergence Approach

Many real-life data sets can be analyzed using Linear Mixed Models (LMMs). Since these are ordinarily based on normality assumptions, under small deviations from the model the inference can be highly unstable when the associated parameters are estimated by classical methods. On the other hand, the density power divergence (DPD) family, which measures the discrepancy between two probability density functions, has been successfully used to build robust estimators with high stability associated with minimal loss in efficiency. Here, we develop the minimum DPD estimator (MDPDE) for independent but non identically distributed observations in LMMs. We prove the theoretical properties, including consistency and asymptotic normality. The influence function and sensitivity measures are studied to explore the robustness properties. As a data based choice of the MDPDE tuning parameter $α$ is very important, we propose two candidates as "optimal" choices, where optimality is in the sense of choosing the strongest downweighting that is necessary for the particular data set. We conduct a simulation study comparing the proposed MDPDE, for different values of $α$, with the S-estimators, M-estimators and the classical maximum likelihood estimator, considering different levels of contamination. Finally, we illustrate the performance of our proposal on a real-data example.

stat.ME↗

arXiv2020

Robust Hypothesis Testing and Model Selection for Parametric Proportional Hazard Regression Models

The semi-parametric Cox proportional hazards regression model has been widely used for many years in several applied sciences. However, a fully parametric proportional hazards model, if appropriately assumed, can often lead to more efficient inference. To tackle the extreme non-robustness of the traditional maximum likelihood estimator in the presence of outliers in the data under such fully parametric proportional hazard models, a robust estimation procedure has recently been proposed extending the concept of the minimum density power divergence estimator (MDPDE) under this set-up. In this paper, we consider the problem of statistical inference under the parametric proportional hazards model and develop robust Wald-type hypothesis testing and model selection procedures using the MDPDEs. We have also derived the necessary asymptotic results which are used to construct the testing procedure for general composite hypothesis and study its asymptotic powers. The claimed robustness properties are studied theoretically via appropriate influence function analyses. We have studied the finite sample level and power of the proposed MDPDE based Wald type test through extensive simulations where comparisons are also made with the existing semi-parametric methods. The important issue of the selection of appropriate robustness tuning parameter is also discussed. The practical usefulness of the proposed robust testing and model selection procedures is finally illustrated through three interesting real data examples.

stat.ME↗

arXiv2020

General Robust Bayes Pseudo-Posterior: Exponential Convergence results with Applications

Although Bayesian inference is an immensely popular paradigm among a large segment of scientists including statisticians, most applications consider objective priors and need critical investigations (Efron, 2013, Science). While it has several optimal properties, a major drawback of Bayesian inference is the lack of robustness against data contamination and model misspecification, which becomes pernicious in the use of objective priors. This paper presents the general formulation of a Bayes pseudo-posterior distribution yielding robust inference. Exponential convergence results related to the new pseudo-posterior and the corresponding Bayes estimators are established under the general parametric set-up and illustrations are provided for the independent stationary as well as non-homogeneous models. Several additional details and properties of the procedure are described, including the estimation under fixed-design regression models.

math.ST↗

arXiv2020

On regularization methods based on Rényi's pseudodistances for sparse high-dimensional linear regression models

Several regularization methods have been considered over the last decade for sparse high-dimensional linear regression models, but the most common ones use the least square (quadratic) or likelihood loss and hence are not robust against data contamination. Some authors have overcome the problem of non-robustness by considering suitable loss function based on divergence measures (e.g., density power divergence, gamma-divergence, etc.) instead of the quadratic loss. In this paper we shall consider a loss function based on the Rényi's pseudodistance jointly with non-concave penalties in order to simultaneously perform variable selection and get robust estimators of the parameters in a high-dimensional linear regression model of non-polynomial dimensionality. The desired oracle properties of our proposed method are derived theoretically and its usefulness is illustustrated numerically through simulations and real data examples.

math.ST↗